Escolar Documentos
Profissional Documentos
Cultura Documentos
`Table of Contents
Table of Contents
March 2009
Table of Contents
Executive Summary ............................................................................. 2
Introduction .......................................................................................... 4
TCO Model Framework and Assumptions ......................................... 6
Network Architecture Assumptions .................................................................... 6
Carrier Ethernet Traffic Projections.................................................................... 8
Comparison between Connection Oriented Ethernet (PBB-TE) and MPLS-TE
Architectures ...................................................................................................... 8
Executive Summary
Service providers around the world have chosen Carrier Ethernet technology for next-generation
transport. Ethernet is attractive because it is:
Ubiquitous
Cost-effective
Compatible with IP packet networks
Capable of supporting port speeds from 10 Mbps to 10 Gbps
Service providers also have recognized the need for traffic engineering and resource management in
Carrier Ethernet transport networks. Traffic engineering optimizes the allocation of network
resources, resulting in efficient network resource utilization and therefore creating significant CapEx
and OpEx savings as demonstrated in two recent Network Strategy Partners’ whitepapers1.
There are debates within the industry regarding Carrier Ethernet network architecture. This paper
compares the total cost of ownership (TCO) of two alternative architectures:
1. Connection Oriented Ethernet (COE) with Gridpoint’s E-TERM
2. MPLS-TE
For this analysis, Provider Backbone Bridging – Traffic Engineering (PBB-TE) is the assumed COE
technology used. PBB-TE is part of a developing set of IEEE standards (IEEE 802.1ah, and
802.1Qay) that uses Ethernet-switching hardware as transport elements and an out-of-band
management system that establishes traffic-engineered tunnel resources and service connections for
Ethernet frame-forwarding. PBB-TE also has the advantage of having standardized Ethernet OAM
through the IEEE 802.1ag & ITU Y.1731 standard that supports OAM capabilities similar to
SONET/SDH. This study assumes that Gridpoint’s E-TERM is used as the traffic-engineering
resource management system that controls the PBB-TE Ethernet transport elements.
Multiprotocol Label Switching – Traffic Engineering (MPLS-TE) is a Layer 3 routing technology
that uses a series of distributed routing protocols to set up Label Switched Paths (LSPs). LSPs are
used to establish traffic-engineered service resources for packet forwarding. MPLS-TE provides
basic OAM capabilities for detecting connectivity using Bidirectional Forwarding Detection and for
tracing paths using LSP traceroute.
Both PBB-TE and MPLS-TE are being proposed for use as an Ethernet transport resource layer for
Carrier Ethernet; however, they have very different architectures and cost structures. The cost of
PBB-TE systems can be lower than the cost of MPLS systems due to the MPLS-TE requirement to
implement a more complex series of routing and forwarding hardware capabilities and complex
resiliency frameworks. The lower cost of PBB-TE systems also stems from the PBB-TE path
routing complexity associated with traffic engineering, as it is managed by an out-of-band network
management system (Gridpoint E-TERM).
An Analysis of the Financial Benefits of Traffic Engineering and Traffic Management in Carrier Ethernet Networks:
1
http://0299d3f.netsolhost.com/NewPages/GP1.pdf.
An Analysis of the Financial Benefits of Traffic Engineering and Traffic Management in Wholesale Carrier Ethernet
Networks: http://0299d3f.netsolhost.com/NewPages/GP3.pdf.
PBB-TE also has lower operational expenses than MPLS-TE. One reason for this is that PBB-TE
uses a transport paradigm similar to SONET/SDH, which is familiar to transport department
technicians. Provisioning and network care procedures are similar in nature to those in a
SONET/SDH network. MPLS-TE Carrier Ethernet switches are essentially IP routers; therefore,
provisioning and network care procedures for MPLS-TE networks are equivalent to those of IP
router networks. Routers use a distributed control plane that requires a large staff of highly skilled
and more expensive engineers, while SONET/SDH equipment is relatively simpler. Consequently,
operational expenses for a PBB-TE network are lower than the expenses associated with the MPLS-
TE network.
This document outlines a network model that is representative of a Carrier Ethernet aggregation
network. The model spans a five-year period and performs a TCO analysis that compares the costs
of Connection Oriented Ethernet (PBB-TE) with Gridpoint’s E-TERM and MPLS-TE architectures
for a Tier 1 service provider aggregation network. The results are summarized in Figure 1.
$10
$9
$8
$7
Total Cost of $6 OpEx
Ownership $5 CapEx
($ Millions) $4
$3
$2
$1
$0
COE MPLS-TE
Figure 1. Comparison of the TCO of PBB-TE and MPLS-TE Over a Five-year Period
This analysis shows a 43% savings for Connection Oriented Ethernet (PBB-TE), when compared
with MPLS-TE, in cumulative capital and operations expenses over a five-year period. It should be
noted that most Tier 1 service providers have multiple aggregation networks in each metro area, so
the numbers in this study should be multiplied by the total number of aggregation networks in the
service provider’s network footprint.
Introduction
For service providers to successfully and competitively deliver today’s new packet-based multimedia
services, a replacement for existing SONET/SDH infrastructures is required. A packet-based
transport technology based on Carrier Ethernet is the choice for this new infrastructure. To get this
packet-based infrastructure to provide the deterministic behavior of SONET/SDH, many service
providers agree that traffic engineering and management functionality is required. However, some
service providers are grappling with the question of whether to use Connection Oriented Ethernet
(PBB-TE) or Multiprotocol Label Switching – Traffic Engineering (MPLS-TE) as the basis of the
next-generation Carrier Ethernet transport network. This whitepaper compares the TCO of both
these alternatives:
Connection Oriented Ethernet (PBB-TE) with Gridpoint’s E-TERM for traffic engineering
and network resource allocation
MPLS-TE
Provider Backbone Bridge Traffic Engineering (PBB-TE) adapts Ethernet technology to packet
transport networks. It is based on layered VLAN tags and MAC-in-MAC encapsulation as defined in
IEEE 802.1ah, Provider Backbone Bridging (PBB). PBB-TE, however, differs from PBB by
eliminating MAC address flooding, MAC address learning, and the spanning tree protocol. PBB-TE
uses a central network management system to statically update all Ethernet MAC layer forwarding
tables as depicted in Figure 2. The analysis assumes that Gridpoint’s E-TERM is used to control
Ethernet PBB-TE Ethernet frame-forwarding and traffic engineering in the Carrier Ethernet
network.
Gridpoint Vendor
E-TERM NMS
10 GbE Links
DWDM Transport Internet
MPLS
Regional
Network
Hub CO
PBB-TE's connection-oriented features, traffic engineering capabilities, and its OAM approach are
inspired by SDH/SONET. Compared to Ethernet predecessor’s technologies, PBB-TE has been
designed to behave more predictably, and its behavior can be more easily managed and defined by
the network operator. PBB-TE also implements transport path monitoring and control using
operational, administration, and maintenance frames (OAM) based on the IEEE 802.1ag and
augmented by the ITU Y.1731 standard. PBB-TE can also provide path protection capabilities
similar to the 1:1 unidirectional protection in SDH/SONET networks. As such, PBB-TE is
designed to integrate with service provider transport processes and OSS systems.
MPLS-TE provides a mechanism to create traffic-engineered connection-oriented paths, named
Label Switched Paths (LSPs), between IP routers. Before a path can be calculated, a view of the
network resources is determined using Layer 3 routing protocols, such as Open Shortest Path First
(OSPF) or Intermediate System to Intermediate System (IS-IS), with traffic engineering extensions
(see Figure 3). Based on this network view, a path is selected for an LSP using Constrained Short
Path First (CSPF) algorithm, fulfilling several constraints (bandwidth, end-to-end delay, number of
links traversed, etc.) simultaneously. Once the path has been calculated, the path is signaled using
RSVP-TE or CR-LDP.
Routing Protocols
• CR‐LDP
• RSVP TE
• OSPF
• ISIS
• BGP
MPLS Ethernet
Carrier Ethernet
Aggregation Aggregation Switch
Network 10 GbE Links
DWDM Transport Internet
MPLS
Regional
Network
Hub CO
Given its distributed nature, the MPLS-TE is a complex technology to deploy and maintain within a
network. While it is a good approach for engineering core IP routing networks, it is not the optimal
technology for designing Carrier Ethernet Metro access/aggregation networks. The Metro
access/aggregation network is responsible for backhauling Ethernet traffic from the access to the
metro core or core network. This network should provide cost-effective transport that fits into a
service provider’s transport paradigm. A Connection Oriented Ethernet (PBB-TE) enabled
architecture using the Gridpoint E-TERM is a more cost-effective approach to building the
Ethernet aggregation network for two main reasons:
PBB-TE switching systems leverages existing Ethernet switching technology with a lower
cost structure than MPLS-TE routing/switching hardware
PBB-TE is similar to SDH/SONET and therefore simpler and less expensive to operate
than an MPLS-TE IP routing network
The following sections of this paper present the assumptions and the TCO model results and
explain why Connection Oriented Ethernet (PBB-TE) with Gridpoint’s E-TERM is a more cost-
effective solution than MPLS-TE.
Medium
CO
Medium Small
CO CO
Large
CO
Hub
Regional Network
Metro Aggregation Network CO MPLS Network
Carrier Ethernet & DWDM
Metro Ring
Medium
CO
Large
CO
Small
CO
Small
CO
8 COs on
Ring
Figure 4. Tier 1 Service Provider Metro Aggregation Ring Consisting of Small, Medium,
and Large COs
The architecture and physical topology of the Carrier Ethernet aggregation ring is specified in Figure
2. A DWDM metro aggregation ring combined with Carrier Ethernet switching infrastructure is
used for packet transport. For large and medium COs, a standalone switch provides 1 GbE
interfaces to CO equipment and connects to the DWDM transport using 10 GbE. For small COs, a
Carrier Ethernet blade is integrated into the DWDM transport and provides 1 GbE interfaces to
CO equipment.
The logical Carrier Ethernet network is represented in Figure 5. Ethernet switches and blades are
connected to the Hub CO using a 10-GbE hub and spoke topology over the DWDM ring. The Hub
CO is the point of interconnection with the regional MPLS routing network and IP service edge
routers. The analysis focuses on comparing PBB-TE with MPLS-TE in the aggregation network.
The assumption is that the core regional network uses large carrier class IP routers running MPLS.
10 GbE Links
Internet
Internet
MPLS
DWDM Aggregation Ring Regional
Network
Hub CO
edge of the network. This edge equipment is based on new PBB-TE technology and, therefore, the
equipment cost is higher than that of existing VLAN Ethernet edge technology. The capital costs of
these devices will go down over time; however, to be conservative, higher pricing estimates are used
for the duration of the TCO model analysis not only to reflect today’s market reality but to also find
a lower bound for the PBB-TE cost savings.
The MPLS-TE alternative uses equipment from a leading Carrier Ethernet MPLS vendor for the
capital expense estimate. These cost estimates are based on current equipment pricing.
$10
$9
$8
$7
Total Cost of $6 OpEx
Ownership $5 CapEx
($ Millions) $4
$3
$2
$1
$0
COE MPLS-TE
The cumulative capital expenses are presented in Figure 7. Network equipment is broken down by
access switches, aggregation switches, and hub switches. The Connection Oriented Ethernet (PBB-
TE) alternative is less capital-intensive than MPLS-TE because Layer 2 Connection Oriented
Ethernet (PBB-TE) systems cost less than Layer 2.5 MPLS switches systems.
$4.0
$3.5
$3.0
$2.5
CapEx Hub
($ Millions) $2.0
$1.5 Aggregation
$1.0 Access
$0.5
$0.0
COE MPLS-TE
A breakdown of operations expenses over the five-year period is presented in Figure 8, and a
definition of the operations expense categories is presented in Table 2. From the analysis, two
primary reasons why the OpEx of the Connection Oriented Ethernet (PBB-TE/Gridpoint
E-TERM) solution costs less than MPLS-TE are as follows:
1. The Connection Oriented Ethernet (PBB-TE/Gridpoint E-TERM) architecture is simpler
to operate and therefore incurs lower labor costs than the MPLS-TE architecture. This is
demonstrated in Figure 8 (see “Training,” “Test and Certification Operations,” and
“Network Care”).
2. Vendor service contracts are annual expenses calculated as a percentage of cumulative
CapEx; MPLS-TE has higher CapEx, making its service expenses also higher than the
Connection Oriented Ethernet (PBB-TE/Gridpoint E-TERM) solution. This is
demonstrated in Figure 8 (see “Service Contracts” and “Sparing Costs”).
Cooling Cost
Power Cost
Sparing Costs
Service Contracts
Training
Network Care
Capacity Management
MPLS-TE COE
Conclusion
This paper presents a detailed cost comparison of two alternative approaches to building a Carrier
Ethernet aggregation network with traffic engineering capabilities:
Connection Oriented Ethernet (PBB-TE) with Gridpoint E-TERM
MPLS-TE
The model of a representative Carrier Ethernet aggregation network over a five-year period
demonstrates that the Connection Oriented Ethernet (PBB-TE/Gridpoint E-TERM) alternative is
43% less expensive than the MPLS-TE alternative. The cost savings are a direct result of lower
equipment costs for Connection Oriented Ethernet (PBB-TE) and the lower operational costs of a
Connection Oriented Ethernet (PBB-TE/Gridpoint E-TERM) network.